Monthly Register — September 2025

September closed without a show, a dated clause or a funding print. Twelve months of this column, and the issue that happens to carry the anniversary is the thinnest of the lot — which is either bad luck or a fair description of how the industry spends a September. Design offices come back from leave. Tender desks catch up. Site carries on as though nothing had paused.

So there is no lead this month, and it is better to say so than to dress a trade rumour up as one. What the year is good for is the residue: the arguments that kept coming back, fortnight after fortnight, until they stopped being news and became the weather.

What moved

Retrieval, first, because it never left. The expensive object on a live job is still the document nobody can find, or the sheet everybody can find in three states across four systems. Rework rides on the back of it. When a gang builds from last week's export, the instinct in the site meeting is to talk about skills; the honest reading is usually that nobody ever told them which set was current, and the set they had was the one that arrived by email on Friday afternoon.

Second, sprawl. A mid-size job still runs a model platform, a document platform, a field app, a commercial system and a long tail of email. Forty systems and no single source of truth was the autumn 2024 diagnosis, and it was still the diagnosis in late 2025. Adding a login does not retire the unofficial PDF set. Nor does moving the inbox into the cloud stop people treating email as the contract record — if anything it makes them more comfortable about doing it.

Third, the common data environment as a claim. Everybody has one, to hear the tender replies. A good many of them are a shared drive with permissions, or a hosted warehouse that keeps every file and cannot tell you what state anything is in, or who received it. ISO 19650 stayed the language of the BIM execution plan all year. Suitability codes turned up dutifully on title blocks. The container discipline underneath — work in progress, shared, published, archive, with one published set — travelled a good deal more slowly than the vocabulary did.

Fourth, the ordinary accident. Building from a superseded drawing is still how a job spends money it will later book as a site error. Revision control and version control are still used as though they were the same act, usually by people who would explain the difference correctly if you asked them over a coffee. A folder structure is still mistaken for the product. Meanwhile the metadata that actually lets a later reader reconstruct anything — revision, suitability, transmittal, the party the container went to — sits with a document controller who is, in my experience, the highest-leverage hire on the job and the one most often priced as clerical.

Fifth, the duties that assume a live file. In the United Kingdom the golden thread stayed a record duty rather than a conference word. Higher-risk buildings needed a set that was identifiable, available and change-controlled, not a folder handed over at practical completion. Health and safety information, including the file the Construction (Design and Management) Regulations already name, was compiled late far more often than it was kept current. A paper permit to work fails in the same place a toolbox talk fails: nobody in the room can say which container the instruction came out of.

Sixth, coordination that never reaches a decision, and then the field. Clash detection went on producing lists nobody closed. The site team held the truth in a cabin pack, the office held the system, and the two rarely spoke to each other. By late summer it had become hard to describe the unofficial second set as a vice. It is what a foreman does when the official set cannot be used by a gloved hand in the rain.

Split those six the way the weekly spine splits them and you have most of the year. What late 2025 added was a cover, not a habit. Artificial intelligence became an ordinary sentence on live jobs — a chat window over last week's export, a drawing read into a summary, a hope that the register would reconcile itself overnight. The test did not change with it. Does the tool read a register, or does it wrap a folder? Can it tell you which sheets and which model state make up the current issued set, who received them and for what purpose, or is it summarising whatever happened to be in the download? Openness got promised in the same breath, and a published container that a party outside the authoring licence can actually open is still rarer than a slide with the word open on it. Gulf permit packs still arrived as a drawing set with a model export and a new first page. British golden-thread folders still failed the produce-it-now test more often than anyone admits in a board pack. Photographs and scans piled up as though volume were a record. Labour on site still plotted from a USB.

A year, then, rather than a week. The industry learned to say information management fluently. The issued set stayed a clerical job unless somebody could prove otherwise.

What changed in the standards

Nothing in September that will survive being cited. An archive's birthday is not a standards event, and no new part of ISO 19650 was published this month.

What the year did put on the shelf has been recorded in earlier issues and does not get a September commencement to fill the space. Part 6, on health and safety information, was the January text; it sets out concepts and principles for classifying, sharing and delivering structured health and safety information across the project and asset life cycle. Most BIM execution plans had still not absorbed it by the last working day of September. That is a fact about adoption rather than a scoop about a clause.

The UK guidance layer that used to travel as the UK BIM Framework was restated earlier in the year under the Information Management Initiative. That was not a new ISO part, not a statutory instrument, and it did not rewrite the published 19650 texts that execution plans already cite. It matters here as context: the British conversation spent the year saying information management out loud while the working clauses stayed exactly where they were.

On the open-schema side, the public conversation in the first half of the year concerned tracks of work with ISO on the IFC series — modularisation, and a next generation that would not orphan current deliveries. None of that is a finished schema a document controller could accept in September, and a slide announcing that the next IFC has arrived is not something you can put in an information requirement.

The Building Safety Act's information duties in England were already in force; they did not arrive in September. Gulf permit-stage BIM requirements remain a class of public-record rule wherever the text is public, with owners of their own. As for the wider revision of the series, drafting and comment collection carried on, which is the honest weather: comments, not a published replacement. If a national annex, circular or ballot did move in September, it belongs in a later correction, once it can be cited properly.

One thing to watch

Whether the same six arguments are still occupying the same conversations in October. Who can name the issued set. Whether the common data environment behaves like a state machine or a warehouse with a login. Whether the golden-thread folder can be produced on the day it is asked for. Whether the permit model is the published set or a PDF with a new cover. Whether the site's unofficial pack has retired, or merely bought a better camera.

Watch the word record wherever it appears in a demonstration. If it turns out to mean a dashboard sitting on the same native files, what has been bought is a view. If the assistant cannot tell you revision and suitability, it is a cover sheet with a language model attached. And if three applications now claim to be the environment while the plotter still takes a USB, the year did not move.

One thing worth asking for before the next stage is appointed, and it costs nothing but somebody's afternoon: ask whoever is proposing the environment to name the current published container in front of you — sheets, model state, revision, recipients, purpose — on a job that is already running, not on a demonstration set. Most of what is described above surfaces in the first ten minutes, and the ten minutes are cheaper before the appointment than after it.

It is the question the last twelve months keep circling back to, and it is still the one worth putting to a project on a bad Friday: if the argument started tomorrow, who could show what was actually issued, and who would spend a fortnight reconstructing it from a box?